In the spring/summer of 1997 I was in the market for about 40 workstations for running engineering apps. We were an HP shop at the time and our friendly HP salesperson of the month explained to me how the C180 system boards were pin compatible with the Merced, which was due out in about a year. This would be an investment protection because we all know how fast Intel's first attempt at a 64-bit chip is going to be, right?
When I asked him what the current status of the chip was, he told me that one of the big three auto makers (I think he said GM) had converted almost entirely to HP workstations running beta merceds and the performance was quite good compared to the PA7800 (or did the C180 come with the PA8000?).
The following year we bought Ultra 10's. They cost 1/3 as much and performed just as well as whatever HP was hocking at the time. Rather ironically, the Ultra 10's did actually come with a certain amount of x86 compatiblity with the SunPCi cards that they came with.:)
Sun workstations are very common as desktops for a wide variety of engineering disciplines. Other platforms have made some inroads into this market but the success stories are limited.
NT has made some progress, but the ports of the applications tend to be second-rate ports of the UNIX version. For example, in the late 90's some electrical engineering apps that were ported to Windows required an X server to be run on Windows. At one biomedical engineering firm that I interviewed at, the engineers tried to use NT, but found that it crashed too frequently during multi-day simulations. There are exceptions to this. Autocad used to run best on UNIX. Since the Pentium came out, Autocad shifted to being a Windows-only app.
Until the last year or so the front-ends of commercial engineering apps on Linux were all but unheard of. Some articles that pop up from time to time in EE Times suggest that vendors are starting to take the Linux desktop seriously. Certainly the motion picture industry has taken this route as of late.
Ximian's packaging of GNOME for Linux and Solaris makes it possible for me to deploy Linux and Sun desktops side by side with a bare minimum of duplicate training. I can't wait to deploy Solaris 9 so that I can stop answering olwm questions.
My guess is that when Sun releases their first Linux boxes, they will have ld, libc, the kernel, etc., configured to run Solaris binaries. Shortly afterwards, expect to see some Solaris kernel, linker, and misc. library patches to make Solaris run Sparc Linux binaries as well.
Why go to the trouble? They have spent way to much time and money marketing two points:
Solaris is the big brother of Linux. To a large extent it is: for now Solaris works much better on all but the bottom tier of Sun's servers.
For binary compatibility across generations of hardware and operating systems, you can count on Sun.
That second point falls down a bit between OS revs for many programs of any sort of complexity, but to a large extent, it is true. That significantly eases the headaches of administering and programming on Sun's platform.
I look forward to the day that I can get decent hardware that everyone (in my Sun-centric IT department) is already familiar with that comes with source code that I can do something with.
The Windows kernel and core libraries were compiled with a faster compiler than the Linux Kernel and core libraries. I would like to see the results of an Windows system built with the intel compiler and a Linux system built with the same compiler. At that time, you can start to compare which OS's architecture is better for the benchmark that is being run.
Let's take a look at who can line pockets. You say the senator is from Virginia, right?
AOL lists 109 open jobs in Dulles, Reston, Manassas, etc., VA
A google search for Herndon, VA reveals the following in the first couple pages. Surely some of these companies would like to see lots of people spending a lot of time on the Internet (or least finding significant value in it).
Foundry Networks
12801 Worldgate Drive, 6th Floor
Herndon, VA 20170
Directions to Inktomi in Herndon, VA
Monument One
1Plaza Ridge I, Suite 100
2975 Worldgate DPlaza Ridge I, Suite 100
Herndon, VA 20170
From
Road Runner's jobs page
Accounting Department (Herndon, VA)
Administration Department (Herndon, VA)
Commercial Services Department (Herndon, VA)
Broadband Technology Department (Herndon, VA)
Operations Department (various locations)
Such a dilemma... Sony is cracking down on a cool hack for the Aibo just about the same time that they are coming out with a US version of Linux for the PS2.
Perhaps those that really care about this will give the letter a close read and come up with intelligent points on how they are being unjust or mean. Let them know that because of this you are not going to buy a PS2 or the Linux addon for it. Send these reasons off to both arms of Sony so that they realize that the behavior of one impacts the profits of the other.
Personally, I think Sony would buy a lot here by spending less on lawyers and spending a little on an engineer to help this guy come up with an approved SDK.
To say that because there is no support built into the product means that you have to hire employees is just plain wrong. Anyone that has worked in an environment with a large budget for IT knows that there are system integrators, consultants, and other organizations that will provide hourly support rates for most any software that you can find.
For example, Kerberos is open source right? Go find a company that does security consulting and ask about support. Depending on the company, that support can go anywhere from phone support to a complete integration of Kerberos in the company's mix of open source and proprietary systems.
This is actually a strength of open source. Since multiple consultants, integrators, and just plain talented individuals have access to the code, it is much harder to get locked into one vendor. I am hard pressed to find someone to go to for Solaris support that can issue me source code patches.
Often times machines at universities are not administered by competent IT staff. It is quite common for a professor to have a grad student manage his/her machines. Normally that grad student doesn't really give a damn about anything other than going to school. Most grad student/admins that I ran across did so because their major professor told them to, not because they wanted. They all eagerly waited for a new grad student to come in to take over IT for the research group.
Now couple that with universities' famous reputation for fighting firewalls tooth and nail. What a wonderful combination.
In large enterprises with tens of thousands of users at various locations around the globe, it would be desirable to have a calendar server at each location, since most people are just going to be scheduling things that are relevant to that location. This cuts down on trans-Atlantic bandwidth usage for "doctor appointment at 3:00" appointments.
The geographically dispersed calendar servers should be able to communicate with each other for the occassions that someone in Paris is arranging a meeting in New York with people from the Chicago and San Francisco locations.
Notifications of events, appointments, etc., should be able to be done from the server, not from clients. That is, the server triggers events on clients (PDA's, pagers, some random unmapped X application). The notification mechanism should allow for local plugins to be provided. Shared library plugins are great, but one should allow for simply calling a shell script too.
Time zones should be handled gracefully. Do not assume that all users are in the same timezone. Assume that when people travel, some change their watch and some don't. (Allow for temporary time zones for individuals?)
Don't require that the calendar gets each users' info added into its own database. Be able to interface with directory services such as LDAP, and X.500. Notes, NDS and ActiveDirectory support would be great, but should be secondary to open directory protocols, since the proprietary ones tend to be able to provide the same data through LDAP gateways.
Allow for clustering for high-availability and load balancing.
Provide an NDMP (network data management protocol) (see www.ndmp.org for details) interface so that enterprise backup systems can back the calendar server up, while it is online.
If an external database is used, abstract it so that the database preferred by the installer can be used. MySQL is a hard sell at a shop that runs Oracle on E10k's.
Authentication via PAM
Limited trust relationships that allow collaboration between different companies.
At my previous job, our netware costs were astronomical compared to our costs for Solaris. The biggest problem was that for netware to be stable and perform well with several hundred users, it required (I think) four dual- and quad-processor boxes. Each of them came from Dell at a price similar to the cost of E450 from Sun. (Actually the Sun was about $2k cheaper). The Sun could handle more memory, more PCI devices, 64-bit, 66-Mhz PCI devices, etc., unlike the PC's.
The 450 in this case had 2 250-MHz processors and 512 MB of RAM. Its load was consistently less than.1, and there was no pressure on memory except during backups.
Sun now has PC Netlink which participates fully in NT domains (PDC, BDC, member server). Initial results at my current job suggest it works quite well. Samba has been working flawlessly on the Sun box, for those that choose to access Unix space from NT. Solaris, PC Netlink, and Samba on the Sun are either free or no-cost. The only things that cost us money are hardware, VERITAS Volume Manager and File System, and Legato Networker.
Conclusion: my department at the University of Wisconsin Madison could have saved about $75k on hardware alone by going with a purely Sun solution. When you add in annual per-client Netware maintenance costs, you start to see an even greater savings. Oh, and I guess at that point you can ditch the Netware admin too (keep the UNIX and NT admins though).
The NDA's that companies have asked me to sign are not nearly as annoying or invasive as background check documents. Applying for a job as a system administrator I have been asked to allow prospective employers to check the following:
Driving record
Credit record
Criminal record
My general nature by interviewing those that I associate with, including my neighbors
Medical record
Drug tests
Essentially, I have been asked to allow them to do the same investigation that the FBI did on me for a security clearance in the military. (They did not, however, ask me if I was gay.) In most cases I have found the requests to be quite unreasonable for the job that I am applying for.
What the fsck does my driving record have to do with a desk job that requires less than 5% travel?
What right does a company that proudly loses millions of dollars per year have to my personal financial records? Why does my employer need to know that I broke my arm when I was 12?
The problems you mention with "not working well with daemons" is likely related to the fact that it uses Kerberos IV. If the daemon needs to have more access to AFS directories than you are willing to give to any other user on the system, there is a lot of work to do.
Specifically, you need to stash a password away such that the daemon can authenticate and periodically reauthenticate so that it does not lose the rights that it has.
AFS does allow you to have ACL's based on IP address. As such, if you are running a daemon on a machine than only system administrators have access to, it may not be a big deal to allow everyone on that machine to write to a directory. Other machines, though, may have read-only or no access to the directory.
NFS 4 will have the same problem, as a requirement for it is that Kerberos V is supported as an authentication mechanism. If you don't give world write to a file/directory, then you cannot write to it without a kerberos V ticket.
Not everyone at IBM has caught the open source fever-- please don't assume that all code sent to IBM becomes open source. As an example, take a look at what has happened with AFS. Back in the early to mid 90's, IBM bought Transarc thinking that there was going to be a big push for globally distributed file systems.
IBM recognized that AFS had some deficiencies and pushed forward with DCE/DFS. At first, they gave an (official, I think) EOL on AFS along with the migration path to DCE/DFS. Once AFS customers figured out what they needed to do with DCE/DFS to make it work as well as AFS already worked, they cried foul and encouraged IBM to alter its product plans.
IBM changed their tune, saying that AFS would be supported in the future, and AFS 3.5 came out on short order. It improved file server performance by over 3 times! Meanwhile, DCE has been open sourced (or something similar, I forget) by the OSF and IBM seems to have forgotten about DFS after their big '96 olympics web site replication with DFS was met with a big "so?".
As of late, though, IBM has determined that there is not money in distributed file systems. They have shifted their "development" to India, and declared AFS to be in "maintenance only mode".
[Thread 1][Thread 2]
As such, AFS currently uses Kerberos IV. Pretty much every site that is serious about AFS has licensed the source and patched it for Kerberos V. This patch has been going around for at least 5 years, yet IBM has been unable to integrate it. Even worse yet, look how they have approached security bugs in Linux 2.2.14.
It would make a lot of sense for AFS to be open sourced, but they are making no moves in that direction. I suspect that a similar fate awaits NDS, should IBM get their hands on it. Then again, maybe the leftovers from Transarc and Novell could get together and come up with a Kerberos V implementation for both products.
Computers were designed for programmers, scientists, and engineers. They work for them rather well. They were not designed for kids and adults that act like kids.
Of course, since computers were originally designed, they have evolved. Unix is evolving to meet the needs of today's scientists, gamers, PHB's, and clothing store clerks.
The intelligent thing to do is to select the right tool for the job. Don't select a tool before it is ready (or you are ready) unless you have the ability to change it or the discipline to offer only constructive criticism.
Relying on RAM to perform disk caching is great if you spend your days sitting around reading/. If, however, you commonly read and write to more disk space than it is feasible to have RAM, a whole bunch of RAM does not really help.
For instance, when you are streaming data to disk at a high rate, you really do not want it to hit the file system buffers. Suppose you have a workstation that is writing out a 100 gig of data. Suppose you are incredibly fortunate and have 4 gig of RAM on this machine. You do not want to use up all 4 gig of RAM 25 times as you are writing this data. If you do that, not only are you thrashing your RAM, but you are wasting lots of CPU cycles trying to figure out which buffers you should not use. Rather, you want the data to bypass the buffer cache so that your/etc/resolv.conf lib*.so, and half of/usr/bin stay in FS buffers. See the "Veritas Performance" section in this paper for an intro to Discovered Direct I/O.
Developers and users have a habit of looking at the falling price of storage and the increasing capacity as reasons that they can bloat software, use larger data sets in simulations, etc. The problem with this is that if capacity continues to increase faster than performance increases, the net result is that it takes longer for me to accomplish a similar task as time goes on. I can honestly say that my 386 with a WD 80 MB drive and 2 MB of RAM started WordPerfect 5.1 much quicker than either of my current Pentium PRO machines can load versions of WordPerfect or MS Word that were state of the art when my PPro machines were new.
"Some of the specific changes that they did were actually things I suggested on the Kerberos mailing list," he says. "So I don't know what sort of claims they are trying to make on this."
Doesn't this imply that MS has no right to claim the "enhancements" as their own intellectual property if they were actually publicly displayed before MS made them available in their own version of Kerberos?
But to get prepared for a large DDOS attack, you need to set up clients on 1000's of machines, with most of them staying up. If you have a choice between breaking installing DDOS agents on machines that are reliable, or ones that are unreliable, which would you choose? Why install 2000 DDOS clients if only 1000 of them are going to be up when you launch your attack?
I tend to agree that under some circumstances my ultra 10/440 with a 7200 RPM IDE disk is a dog. As an example, the other day I was pulling data from the CD (on the second IDE controller) to the hard drive (on the first IDE controller). I felt like I was dialing into a HP 425 through my 1200 baud modem again.
However, when I hooked a Plextor SCSI CD-ROM to it and wrote it off to my 10,000 RPM SCSI drive, I could not even tell that anything was running on my machine.
FWIW, you can get PCI UltraSCSI cards that work perfectly on the Ultra 5's and Ultra 10's for about $80. See my Using CD-R or CD-RW drives on Solaris page for details.
I have looked high and low for something like this in the past to deal with a couple problems that I have had to deal with.
The first problem (already solved, very expensively) was to find a way to set up a secure terminal server to access the console of my sun servers. Using a standard terminal server didn't quite meet my needs 'cause I couldn't find one that supported ssh. The best solution that I could find was to get a SCSI terminal server and attach it onto an old sun. This cost me over $1000 and an old sun. If I could have gotten one of these boxes with 16 serial ports, I would have gone with it instead.
The second application that I would use something like this for is a print server. Sure, you can get HP JetDirect cards or standalone print servers from HP, but from my experience they suck pretty bad. Any piece of hardware that makes me walk to it to reboot it more than once per year is begging to be replaced. One of these boxes with a couple parallel ports and an LCD panel to say whose job is being printed to which printer would be awesome.
This is how I found out about it. I even verified it myself by telnetting to the box... maybe I telnetted to port 80, but I think that it was 23. See this article at deja.com for a fairly accurate record.
I think one thing that stifles development is when programmers cannot take ideas or code from work, further develop them, and distribute them outside of work. I certainly have worked on projects for the University of Wisconsin that I think would have benefitted from releasing under an OSS license. Unfortunately, according to some intellectual property lecture I attentended, I need to get approval from some bureaucrat to say that the UW does not stand to lose a chance to make money off of the code I wish to release.
Because of these stupid rules, I have not released much of my code. However, if I create software from home and it is useful at work, it gets released. This is true even if I work on it in off hours specifically for use at work. For smaller projects, I prefer to just do them at home and release the code. Note that I get paid dirt working from home, but I do have fun.
So far, it seems as though RHAT and LNUX have no such stupid rules. Should they come up with stupid rules like that, I think that they would slowly lose their importance in the Linux world.
Slashdot Mirror
In the spring/summer of 1997 I was in the market for about 40 workstations for running engineering apps. We were an HP shop at the time and our friendly HP salesperson of the month explained to me how the C180 system boards were pin compatible with the Merced, which was due out in about a year. This would be an investment protection because we all know how fast Intel's first attempt at a 64-bit chip is going to be, right?
:)
When I asked him what the current status of the chip was, he told me that one of the big three auto makers (I think he said GM) had converted almost entirely to HP workstations running beta merceds and the performance was quite good compared to the PA7800 (or did the C180 come with the PA8000?).
The following year we bought Ultra 10's. They cost 1/3 as much and performed just as well as whatever HP was hocking at the time. Rather ironically, the Ultra 10's did actually come with a certain amount of x86 compatiblity with the SunPCi cards that they came with.
Sun workstations are very common as desktops for a wide variety of engineering disciplines. Other platforms have made some inroads into this market but the success stories are limited.
NT has made some progress, but the ports of the applications tend to be second-rate ports of the UNIX version. For example, in the late 90's some electrical engineering apps that were ported to Windows required an X server to be run on Windows. At one biomedical engineering firm that I interviewed at, the engineers tried to use NT, but found that it crashed too frequently during multi-day simulations. There are exceptions to this. Autocad used to run best on UNIX. Since the Pentium came out, Autocad shifted to being a Windows-only app.
Until the last year or so the front-ends of commercial engineering apps on Linux were all but unheard of. Some articles that pop up from time to time in EE Times suggest that vendors are starting to take the Linux desktop seriously. Certainly the motion picture industry has taken this route as of late.
Ximian's packaging of GNOME for Linux and Solaris makes it possible for me to deploy Linux and Sun desktops side by side with a bare minimum of duplicate training. I can't wait to deploy Solaris 9 so that I can stop answering olwm questions.
My guess is that when Sun releases their first Linux boxes, they will have ld, libc, the kernel, etc., configured to run Solaris binaries. Shortly afterwards, expect to see some Solaris kernel, linker, and misc. library patches to make Solaris run Sparc Linux binaries as well.
Why go to the trouble? They have spent way to much time and money marketing two points:
- Solaris is the big brother of Linux. To a large extent it is: for now Solaris works much better on all but the bottom tier of Sun's servers.
- For binary compatibility across generations of hardware and operating systems, you can count on Sun.
That second point falls down a bit between OS revs for many programs of any sort of complexity, but to a large extent, it is true. That significantly eases the headaches of administering and programming on Sun's platform.I look forward to the day that I can get decent hardware that everyone (in my Sun-centric IT department) is already familiar with that comes with source code that I can do something with.
The Windows kernel and core libraries were compiled with a faster compiler than the Linux Kernel and core libraries. I would like to see the results of an Windows system built with the intel compiler and a Linux system built with the same compiler. At that time, you can start to compare which OS's architecture is better for the benchmark that is being run.
It seems as though there was a company not too long ago that enabled people to share lots of music using peer-to-peer technologies.
Oh wait, if you distribute data using peer-to-peer technology, the copyright holders will get really pissed.
AOL lists 109 open jobs in Dulles, Reston, Manassas, etc., VA
A google search for Herndon, VA reveals the following in the first couple pages. Surely some of these companies would like to see lots of people spending a lot of time on the Internet (or least finding significant value in it).
Foundry Networks
12801 Worldgate Drive, 6th Floor
Herndon, VA 20170
Directions to Inktomi in Herndon, VA
Monument One
1Plaza Ridge I, Suite 100 2975 Worldgate DPlaza Ridge I, Suite 100
Herndon, VA 20170
From Road Runner's jobs page
Accounting Department (Herndon, VA)
Administration Department (Herndon, VA)
Commercial Services Department (Herndon, VA)
Broadband Technology Department (Herndon, VA)
Operations Department (various locations)
Such a dilemma... Sony is cracking down on a cool hack for the Aibo just about the same time that they are coming out with a US version of Linux for the PS2.
Perhaps those that really care about this will give the letter a close read and come up with intelligent points on how they are being unjust or mean. Let them know that because of this you are not going to buy a PS2 or the Linux addon for it. Send these reasons off to both arms of Sony so that they realize that the behavior of one impacts the profits of the other.
Personally, I think Sony would buy a lot here by spending less on lawyers and spending a little on an engineer to help this guy come up with an approved SDK.
For example, Kerberos is open source right? Go find a company that does security consulting and ask about support. Depending on the company, that support can go anywhere from phone support to a complete integration of Kerberos in the company's mix of open source and proprietary systems.
This is actually a strength of open source. Since multiple consultants, integrators, and just plain talented individuals have access to the code, it is much harder to get locked into one vendor. I am hard pressed to find someone to go to for Solaris support that can issue me source code patches.
Often times machines at universities are not administered by competent IT staff. It is quite common for a professor to have a grad student manage his/her machines. Normally that grad student doesn't really give a damn about anything other than going to school. Most grad student/admins that I ran across did so because their major professor told them to, not because they wanted. They all eagerly waited for a new grad student to come in to take over IT for the research group.
Now couple that with universities' famous reputation for fighting firewalls tooth and nail. What a wonderful combination.
The geographically dispersed calendar servers should be able to communicate with each other for the occassions that someone in Paris is arranging a meeting in New York with people from the Chicago and San Francisco locations.
Notifications of events, appointments, etc., should be able to be done from the server, not from clients. That is, the server triggers events on clients (PDA's, pagers, some random unmapped X application). The notification mechanism should allow for local plugins to be provided. Shared library plugins are great, but one should allow for simply calling a shell script too.
Time zones should be handled gracefully. Do not assume that all users are in the same timezone. Assume that when people travel, some change their watch and some don't. (Allow for temporary time zones for individuals?)
Don't require that the calendar gets each users' info added into its own database. Be able to interface with directory services such as LDAP, and X.500. Notes, NDS and ActiveDirectory support would be great, but should be secondary to open directory protocols, since the proprietary ones tend to be able to provide the same data through LDAP gateways.
Allow for clustering for high-availability and load balancing.
Provide an NDMP (network data management protocol) (see www.ndmp.org for details) interface so that enterprise backup systems can back the calendar server up, while it is online.
If an external database is used, abstract it so that the database preferred by the installer can be used. MySQL is a hard sell at a shop that runs Oracle on E10k's.
Authentication via PAM
Limited trust relationships that allow collaboration between different companies.
The 450 in this case had 2 250-MHz processors and 512 MB of RAM. Its load was consistently less than .1, and there was no pressure on memory except during backups.
Sun now has PC Netlink which participates fully in NT domains (PDC, BDC, member server). Initial results at my current job suggest it works quite well. Samba has been working flawlessly on the Sun box, for those that choose to access Unix space from NT. Solaris, PC Netlink, and Samba on the Sun are either free or no-cost. The only things that cost us money are hardware, VERITAS Volume Manager and File System, and Legato Networker.
Conclusion: my department at the University of Wisconsin Madison could have saved about $75k on hardware alone by going with a purely Sun solution. When you add in annual per-client Netware maintenance costs, you start to see an even greater savings. Oh, and I guess at that point you can ditch the Netware admin too (keep the UNIX and NT admins though).
Essentially, I have been asked to allow them to do the same investigation that the FBI did on me for a security clearance in the military. (They did not, however, ask me if I was gay.) In most cases I have found the requests to be quite unreasonable for the job that I am applying for.
What the fsck does my driving record have to do with a desk job that requires less than 5% travel? What right does a company that proudly loses millions of dollars per year have to my personal financial records? Why does my employer need to know that I broke my arm when I was 12?
Specifically, you need to stash a password away such that the daemon can authenticate and periodically reauthenticate so that it does not lose the rights that it has.
AFS does allow you to have ACL's based on IP address. As such, if you are running a daemon on a machine than only system administrators have access to, it may not be a big deal to allow everyone on that machine to write to a directory. Other machines, though, may have read-only or no access to the directory.
NFS 4 will have the same problem, as a requirement for it is that Kerberos V is supported as an authentication mechanism. If you don't give world write to a file/directory, then you cannot write to it without a kerberos V ticket.
IBM recognized that AFS had some deficiencies and pushed forward with DCE/DFS. At first, they gave an (official, I think) EOL on AFS along with the migration path to DCE/DFS. Once AFS customers figured out what they needed to do with DCE/DFS to make it work as well as AFS already worked, they cried foul and encouraged IBM to alter its product plans.
IBM changed their tune, saying that AFS would be supported in the future, and AFS 3.5 came out on short order. It improved file server performance by over 3 times! Meanwhile, DCE has been open sourced (or something similar, I forget) by the OSF and IBM seems to have forgotten about DFS after their big '96 olympics web site replication with DFS was met with a big "so?".
As of late, though, IBM has determined that there is not money in distributed file systems. They have shifted their "development" to India, and declared AFS to be in "maintenance only mode". [Thread 1] [Thread 2]
As such, AFS currently uses Kerberos IV. Pretty much every site that is serious about AFS has licensed the source and patched it for Kerberos V. This patch has been going around for at least 5 years, yet IBM has been unable to integrate it. Even worse yet, look how they have approached security bugs in Linux 2.2.14.
It would make a lot of sense for AFS to be open sourced, but they are making no moves in that direction. I suspect that a similar fate awaits NDS, should IBM get their hands on it. Then again, maybe the leftovers from Transarc and Novell could get together and come up with a Kerberos V implementation for both products.
Of course, since computers were originally designed, they have evolved. Unix is evolving to meet the needs of today's scientists, gamers, PHB's, and clothing store clerks.
The intelligent thing to do is to select the right tool for the job. Don't select a tool before it is ready (or you are ready) unless you have the ability to change it or the discipline to offer only constructive criticism.
For instance, when you are streaming data to disk at a high rate, you really do not want it to hit the file system buffers. Suppose you have a workstation that is writing out a 100 gig of data. Suppose you are incredibly fortunate and have 4 gig of RAM on this machine. You do not want to use up all 4 gig of RAM 25 times as you are writing this data. If you do that, not only are you thrashing your RAM, but you are wasting lots of CPU cycles trying to figure out which buffers you should not use. Rather, you want the data to bypass the buffer cache so that your /etc/resolv.conf lib*.so, and half of /usr/bin stay in FS buffers. See the "Veritas Performance" section in this paper for an intro to Discovered Direct I/O.
Developers and users have a habit of looking at the falling price of storage and the increasing capacity as reasons that they can bloat software, use larger data sets in simulations, etc. The problem with this is that if capacity continues to increase faster than performance increases, the net result is that it takes longer for me to accomplish a similar task as time goes on. I can honestly say that my 386 with a WD 80 MB drive and 2 MB of RAM started WordPerfect 5.1 much quicker than either of my current Pentium PRO machines can load versions of WordPerfect or MS Word that were state of the art when my PPro machines were new.
Doesn't this imply that MS has no right to claim the "enhancements" as their own intellectual property if they were actually publicly displayed before MS made them available in their own version of Kerberos?
Any thoughts about building and distributing --with-rsaref versions of ssh and openssl so that the .edu's can legally make use of these packages?
But to get prepared for a large DDOS attack, you need to set up clients on 1000's of machines, with most of them staying up. If you have a choice between breaking installing DDOS agents on machines that are reliable, or ones that are unreliable, which would you choose? Why install 2000 DDOS clients if only 1000 of them are going to be up when you launch your attack?
However, when I hooked a Plextor SCSI CD-ROM to it and wrote it off to my 10,000 RPM SCSI drive, I could not even tell that anything was running on my machine.
FWIW, you can get PCI UltraSCSI cards that work perfectly on the Ultra 5's and Ultra 10's for about $80. See my Using CD-R or CD-RW drives on Solaris page for details.
The first problem (already solved, very expensively) was to find a way to set up a secure terminal server to access the console of my sun servers. Using a standard terminal server didn't quite meet my needs 'cause I couldn't find one that supported ssh. The best solution that I could find was to get a SCSI terminal server and attach it onto an old sun. This cost me over $1000 and an old sun. If I could have gotten one of these boxes with 16 serial ports, I would have gone with it instead.
The second application that I would use something like this for is a print server. Sure, you can get HP JetDirect cards or standalone print servers from HP, but from my experience they suck pretty bad. Any piece of hardware that makes me walk to it to reboot it more than once per year is begging to be replaced. One of these boxes with a couple parallel ports and an LCD panel to say whose job is being printed to which printer would be awesome.
This is how I found out about it. I even verified it myself by telnetting to the box... maybe I telnetted to port 80, but I think that it was 23. See this article at deja.com for a fairly accurate record.
I think one thing that stifles development is when programmers cannot take ideas or code from work, further develop them, and distribute them outside of work. I certainly have worked on projects for the University of Wisconsin that I think would have benefitted from releasing under an OSS license. Unfortunately, according to some intellectual property lecture I attentended, I need to get approval from some bureaucrat to say that the UW does not stand to lose a chance to make money off of the code I wish to release.
Because of these stupid rules, I have not released much of my code. However, if I create software from home and it is useful at work, it gets released. This is true even if I work on it in off hours specifically for use at work. For smaller projects, I prefer to just do them at home and release the code. Note that I get paid dirt working from home, but I do have fun.
So far, it seems as though RHAT and LNUX have no such stupid rules. Should they come up with stupid rules like that, I think that they would slowly lose their importance in the Linux world.